Tuesday, 2026-04-07

opendevreviewAnishere Mariam proposed openstack/ironic master: Add node history state timestamps  https://review.opendev.org/c/openstack/ironic/+/98255500:00
opendevreviewKaifeng Wang proposed openstack/ironic-tempest-plugin master: Allow removing volume connectors on some states  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/98185901:38
opendevreviewDavid Nwosu proposed openstack/ironic master: Add global node history endpoint  https://review.opendev.org/c/openstack/ironic/+/98314407:50
opendevreviewDavid Nwosu proposed openstack/ironic master: Add global node history endpoint  https://review.opendev.org/c/openstack/ironic/+/98314408:06
opendevreviewAbhishek Bongale proposed openstack/ironic master: Limit interfaces to fake-graphical for testing  https://review.opendev.org/c/openstack/ironic/+/98227008:07
opendevreviewAbhishek Bongale proposed openstack/ironic-tempest-plugin master: Use startswith for interface validation in tests  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/98296008:28
opendevreviewAbhishek Bongale proposed openstack/ironic-tempest-plugin master: Add conditional graphical console validation  https://review.opendev.org/c/openstack/ironic-tempest-plugin/+/98207708:28
VictorAkoh[m]hroy: Hello! I trust you're good. Were you able to check it for me, please?08:29
opendevreviewPierre Riteau proposed openstack/networking-generic-switch stable/2026.1: CI: Remove setuptools pin  https://review.opendev.org/c/openstack/networking-generic-switch/+/98352109:52
victor_obi_TheJulia, JayF: Hi, working on bug #2097102. Reviewers noted that conductor_affinity and _filter_by_conductor have different semantics. Is there a recommended approach to push conductor filtering to the DB layer while preserving hash ring semantics?11:42
TheJuliagood morning everyone13:14
cardoedown in the weeds of devstack13:39
cardoeGotta update our docs or find default config13:39
TheJuliaRegarding victor_obi_'s question, I'm not sure there really is :( It feels like we likely need to quantify and have a bug that we just don't really understand completely becuase of perception bias and we also don't have solid data beyond kaifeng's bug report13:44
TheJuliaAnd I think we might need to change the meaning, slightly in that case, dunno.13:44
opendevreviewJulia Kreger proposed openstack/ironic master: Remove IRONIC_RAMDISK_TYPE devstack setting  https://review.opendev.org/c/openstack/ironic/+/98349013:52
TheJuliacardoe: are you referring to https://review.opendev.org/c/openstack/ironic/+/983490/2/doc/source/contributor/devstack-guide.rst ?14:05
TheJuliaor other issues?14:05
cardoeOther issues. I used the local.conf from there. I removed tinyipa14:05
cardoeSo network segment range plugin isn't enabled by default. I've enabled it. But my vxlan doesn't have an allocation range that it can use.14:09
TheJuliaOH14:18
TheJuliaso you tried vlxan with a skelaton of the settings from the guide?14:18
TheJuliause the CI job local.conf, it will make you happier, so much happier14:18
JayFafter you fix the documented one though, please :D 14:18
JayFI use the local.conf from devstack guide, but usually carry forward a working local.conf for a while14:19
TheJuliawell, its a mixed state, he is pulling in other services which require extra config outside of the base example14:19
JayFI don't have a working one r/n but that doc is how GR-OSS people use it14:19
JayFaha14:19
TheJuliaThere is already a change pending with the docs being fixed inside14:19
JayFyeah more services I can't make promises, only so much [stuff] can fit in a bucket ;) 14:19
TheJuliayeah14:19
TheJuliaThere is now a bare metal bear doing disco in the slide deck14:48
JayFDo you think there's value in having like a 5-minute here's where ironic fits in the stack as an intro to that update before I start emceeing in all of the individual updates?14:56
TheJuliaI think that brings value, yes14:56
TheJuliaI've been on this theme of "over communicate; just don't assume people know what your talking about" kick as of recent14:57
TheJuliaWe likely ought to also point back to launchpad at the very end14:57
opendevreviewPavlo Shchelokovskyy proposed openstack/ironic master: Only label kernel-less ISO as config drive  https://review.opendev.org/c/openstack/ironic/+/98357614:59
JayFYeah it makes sense. I will get some good slides wrapping the start and end today14:59
TheJuliaI'm already on the end15:01
TheJuliaand at least, a little bit of artwork15:01
TheJulia"artwork"15:01
* TheJulia walks the canine overlord15:01
TheJuliaerr, corgi. Its a corgi, corgis are corgis, there is no alternative to corgis15:02
TheJuliaJust so we're all on the same page: JayF is being the host/introducer/closer. alegacy is doing standalone netowrking, clif is doing TBN, I'm doing vxlan, cardoe is doing other additions. Jay do you want to do deprecations as well?15:22
JayFyeah15:33
cardoeI guess I need to flesh out the other additions slides.15:38
clifsounds good to me15:38
cardoeTheJulia, JayF: I'm not pulling in any other services. I literally just tried to use the local.conf from the devstack guide.15:39
cardoeI replaced "ipmi" with "redfish" globally.15:39
JayFthe ironic w/nova config there is the one I usually use15:39
JayFno don't do that15:39
JayFthere's one place to change, with a comment on it, to change to redfish15:39
JayFonly change that15:39
TheJuliause of redfish requires the sushy-tools repo afaik15:39
cardoeI said "globally" I mean I read the comment and did the change the comment said.15:40
JayFack good 15:40
JayFthen IDK why it would boom, I am in meetings all morning but can help if there's a specific error/stuck point15:40
cardoeIt came up just fine but then ate itself on tinyipa.15:40
cardoeWhich Julia is fixing.15:41
JayFah, so IRONIC_RAMDISK_TYPE=dib would fix it too15:41
JayFgood15:41
cardoeI deleted the line and ran ./stack.sh again and it's BOOM15:41
cardoeError while executing command: HttpException: 503, Unable to create the network. No project network is available for allocation.15:41
cardoe++lib/neutron_plugins/services/l3:create_neutron_initial_network:202  oscwrap --os-cloud devstack --os-region RegionOne network create private -f value -c id15:41
JayFyou're going to need to backtrack from there to find why it failed15:42
JayF503 there implies a bad config in a service perhaps15:42
JayFIs this 100% upstream or using some patched repos?15:42
opendevreviewDavid Nwosu proposed openstack/ironic master: Add project field to node history and record deployment transitions  https://review.opendev.org/c/openstack/ironic/+/98334915:43
cardoehttps://www.irccloud.com/pastebin/GqppFXAZ/15:49
cardoeNo patched repos. I literally git cloned devstack. Had this diff (without the tinyipa removal line). Ran ./stack.sh and it was up and all that. I tried to do something but tinyipa was problematic. I mentioned that here. TheJulia gave a patch to remove the tinyipa line. So I removed it locally and ran ./stack.sh again. And I'm getting the 503.15:50
cardoeI've tried to run unstack.sh and then stack.sh again and same issue.15:50
JayFThat config should work, although I'll note unstack/stack with Ironic doesn't always work especially if the local.conf changes or the stack itself was busted the first time15:51
TheJuliait won't work without sushy-tools being pulled in afaik15:51
JayFIf it's broken it's probably a regression; someone fixed something in CI but not in the devstack docs15:51
JayFTheJulia: The code pip-installs sushy-tools from pypi if it's not in required-projects IIRC. If that changed it would break things here.15:51
TheJuliaI'm not sure it really does, but requires taking a look. i.e. its easy to bikeshed this issue and someone just needs to give it as spin/repro/fix15:52
JayFAFAICT there is no reason the config as shown by cardoe shouldn't work which implies to me he may be hitting a local issue or a new bug :( (new bug :CCCC)15:53
TheJuliacardoe: trying to remember but I think that is when the actual allocation fails out for segmenting which makes me think $something else is needed config wise15:53
cardoeThat's what I think. There's a chunk of config missing.15:53
cardoeSo now locally I've edited neutron.conf and added service_plugins += network-segment-range15:54
TheJuliaoh, you need "segments"15:54
TheJuliamaybe?15:54
cardoeWell no. You automagically get segments always.15:54
JayFMy main question would revolve around why is this broken in devstack-local but not CI15:54
JayFit implies that we changed a thing to fix CI, but I don't think we did15:54
TheJuliauhhh, I haven't I don't think15:55
cardoethe segments plugin just enables the ability to query the segments API endpoints15:55
cardoeBut when I enable network-segment-range, there's no available ranges to allocate from.15:55
cardoeWhich is where my 503 is happening.15:55
cardoeI dunno zuul well enough to understand how it's local.conf in CI is being built up.15:58
cardoeI had a VM I built forever ago with devstack and have been using it.15:58
JayFlook for the .local.conf.auto in the CI job 15:58
JayFthat's basically a rendered local.conf that has more detail than you should need locally but could still contain useful bits15:59
cardoeBut I'm a derp and I nuked it. So just trying to build it up again using the docs. I previously used the docs which is why I submitted some fixes a long while back.15:59
JayFyeah I think it's extremely extremely important for that local.conf in docs to work16:00
JayFso anything we can do to make it go would be appreciated16:00
JayFthat's the discovarble "one machine ironic" new developers go to16:00
cardoeThe issue is that "initialize_network_segment_range_support()" is not being called within Neutron.16:02
TheJuliadisjointed network config?16:03
cardoeI'm fairly certain the reason why this worked the first time around is cause the DB didn't exist and when it was initially populated it triggers another code path.16:03
TheJuliadid you run clean to wipe it all out?16:03
cardoeWell they've got like 3 different ways to possibly initialize it.16:03
TheJuliaThat is one thing I've always done16:03
cardoeI can. I just thought that changing a setting and running ./stack.sh should work.16:04
cardoeLike it should just switch me to dib from tinyipa.16:04
TheJuliaIf the destination file names are the same, I think the plugin will skip the download again if the files are already there16:05
cardoeI'm not even getting there again though.16:06
cardoeSomewhere the neutron DB is getting wiped in the ./stack.sh call.16:09
cardoeCause manually telling it to use network-segment-range in /etc/neutron/neutron.conf and systemctl restart devstack@neutron-api.service results in stuff working again.16:09
cardoeBut running ./stack.sh again turns off network-segment-range and the segment allocations DB table is empty again resulting in the 503.16:10
cardoeTesting out 'Q_SERVICE_PLUGIN_CLASSES=network_segment_range' in my local.conf16:15
cardoeWell that got me farther.16:27
cardoe++/opt/stack/ironic/devstack/lib/ironic:create_bridge_and_vms:2577  sudo ip route add 10.0.6.0/24 via 172.24.5.250 dev enp3s0 onlink16:27
JayFI think ./unstack ./stack is hurting you16:27
JayFa clean VM might help16:27
cardoeRTNETLINK answers: File exists16:27
cardoeJayF: it was literally a fresh VM yesterday.16:27
cardoeRan ./stack.sh using the local.conf from the docs. But it used tinyipa so I just tried to switch to dib. So I deleted the line I was told and ran ./stack.sh again.16:28
TheJuliathis is why I always run ./clean.sh16:28
cardoeI'm running ./clean.sh now.16:29
JayFyeah I think that's when things went sideways; it broke the first time and you are trying ont he same VM16:30
JayFany devstack stack failure for me === kill the VM with fire16:30
cardoeStacking again.16:32
cardoeI can. It just shouldn't be that bad.16:33
opendevreviewVictor proposed openstack/ironic master: Fix API pagination limit for conductor filtering  https://review.opendev.org/c/openstack/ironic/+/98344916:33
opendevreviewJulia Kreger proposed openstack/ironic-python-agent master: Migrate package metadata from setup.cfg to pyproject.toml  https://review.opendev.org/c/openstack/ironic-python-agent/+/98286416:34
opendevreviewJulia Kreger proposed openstack/ironic-python-agent master: Update hacking to 7.0.0  https://review.opendev.org/c/openstack/ironic-python-agent/+/98286716:34
cardoesame failure in ip route add... I don't see us ever cleaning that up.16:38
TheJuliathe virtual interface is likely not removed in that case. you can likely manually clean it but removal or recation is kind of dubious since it likely doesn't matter if it already exists16:39
JayFor make the add idempotent16:40
cardoeWell "replace" is the idempotent version of add.16:42
cardoeWe use "replace" in almost all the other calls.16:42
JayFI'm trying to remember if `ip r r` errs if it doesn't exist16:42
JayFif not, you may have a winner16:42
cardoeTheJulia: well it errored on that route even after running ./clean.sh16:45
TheJuliaso yeah, lets fix the call or remove it ?16:46
opendevreviewDavid Nwosu proposed openstack/ironic master: Add global node history endpoint  https://review.opendev.org/c/openstack/ironic/+/98314416:46
* cardoe table flips.16:47
cardoehttps://www.irccloud.com/pastebin/Dz1WdMgF/16:47
cardoeI dunno why it's initializing sometimes and not others.16:48
cardoeThat's after a clean.sh. It's almost like a race condition.16:48
cardoeWhich is mentioned in the driver code...16:48
TheJuliaif tenant_network_type is not set, its geneve by default so not really sure of the entire context of where your at16:48
TheJuliathe plugin otherwise generally works on provider networks as well, the downside is the tenant_network_type takes over for default private network creation in the neutron plugin16:49
cardoeuhh yeah there's some race.16:51
TheJuliaYes, inherent race due to async plugin execution16:51
cardoeneutron-api is populating it... then there's DB errors with MySQL while it's running "init_neutron" and allocations are empty.16:52
cardoeafter the DB errors16:52
TheJuliaugh16:52
cardoerestarting neurton-api makes the neutron API work again.16:53
cardoeSo something is bad touching the DB in the middle.16:53
TheJuliaI encountered this exact same issue at one point.... 16:53
TheJuliauhhhhhh16:53
TheJuliaIt might have been intermittent on my machine16:53
cardoeI'm about to throw out my stupid idea of "this should be repairable" and just start with a fresh VM16:54
JayFI think your idea is solid16:54
JayFyou're just running into the same wall others with the same mind have hit16:54
JayFI don't disagree that stack/unstack/clean should all work forever on a good VM, it's just never really worked more than a little while consistently16:54
cardoeSo I'm just trying to test my migrations for the runbook traits.16:55
cardoeAnd then also try to work out the changes I've got for devstack.16:55
cardoeJayF: I changed the wording on https://review.opendev.org/c/openstack/python-ironicclient/+/981805 btw16:56
cardoehttps://review.opendev.org/c/openstack/devstack/+/983361 is what I'm trying to change for devstack but it's broken on neutron... neutron is why I'm trying to change it.16:57
-opendevstatus- NOTICE: Load on the opendev.org Gitea backends is under control again for now, if any Zuul jobs failed with SSL errors or disconnects reaching the service prior to 16:15 UTC they can be safely rechecked17:03
opendevreviewJay Faulkner proposed openstack/ironic master: Remove iLO hardware type and interfaces  https://review.opendev.org/c/openstack/ironic/+/98270617:12
*** dmellado471 is now known as dmellado4717:48
opendevreviewJulia Kreger proposed openstack/ironic-python-agent master: Enforce TLS 1.2 minimum across all connections  https://review.opendev.org/c/openstack/ironic-python-agent/+/98217418:56
opendevreviewJulia Kreger proposed openstack/ironic-python-agent master: Upgrade TLS certificates to P-384 curve with configurable support  https://review.opendev.org/c/openstack/ironic-python-agent/+/98217519:27
cardoeTheJulia: https://bugs.launchpad.net/neutron/+bug/2114451 I think we're able to do this with the BindLevel right?20:21
TheJuliaI mean, At least with our part of the binding code I I'm unsure it really matters20:40
TheJuliabecause we're not treating segments as spine/leaf subnets20:40
TheJuliawe're treating segments as intermediary networks which are attached at different binding levels20:40
TheJuliaso, does it really matter when its the L2 domain.20:40
opendevreviewVictor proposed openstack/ironic master: Fix API pagination limit for conductor filtering  https://review.opendev.org/c/openstack/ironic/+/98344920:59
cardoeself flogging with devstack complete21:25
cardoechange ip route add -> ip route replace. enable network-segment-range plugin. Are the two changes I needed from our documented local.conf but that's not all21:26
cardoeyou must manually systemctl stop devstack@neutron-api.service and devstack@g-api.service21:26
cardoeWe never teardown that route even with a ./clean.sh21:28
cardoeSo the replace might be wrong-ish.21:29
TheJuliasounds like there is something which doesn't occur in devstack in ci, but occurs locally which may be an extra variable since we're generally not having to restart neutron today21:43
JayFTheJulia: clif: others; I've modified the slides in the ironic oi live -- I think we're good to go. I'm assuming I'm handling the "misc" feature portions and transitioning myself in between all speakers to help draw the through line21:50
cardoeTheJulia: it’s totally a race with neutron. If it’s running and it restarts it then it ends up with the VLAN and VXLAN allocations empty. If it’s stopped first then the wipe happens before it starts it up.22:58
cardoeIt also only happens on my shiny box. My dusty box it doesn’t happen on.22:58
cardoeDusty is still a Zen 5 box.23:00
TheJuliafun!23:05
TheJulia:(23:05

Generated by irclog2html.py 4.1.0 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!